update model checkpoints

README.md

@@ -1,14 +1,11 @@
 ---
-license: gemma
 library_name: peft
 tags:
 - alignment-handbook
-- trl
-- sft
 - generated_from_trainer
-base_model: google/gemma-7b
 datasets:
 - llama-duo/synth_summarize_dataset_dedup
+base_model: google/gemma-7b
 model-index:
 - name: gemma7b-summarize-gpt4o-64k
   results: []
@@ -21,7 +18,7 @@ should probably proofread and complete it, then remove this comment. -->
 
 This model is a fine-tuned version of [google/gemma-7b](https://huggingface.co/google/gemma-7b) on the llama-duo/synth_summarize_dataset_dedup dataset.
 It achieves the following results on the evaluation set:
-- Loss: 5.5381
+- Loss: 2.5157
 
 ## Model description
 
@@ -42,43 +39,38 @@ More information needed
 The following hyperparameters were used during training:
 - learning_rate: 0.0002
 - train_batch_size: 4
-- eval_batch_size: 4
+- eval_batch_size: 2
 - seed: 42
 - distributed_type: multi-GPU
-- num_devices: 2
+- num_devices: 8
 - gradient_accumulation_steps: 2
-- total_train_batch_size: 16
-- total_eval_batch_size: 8
+- total_train_batch_size: 64
+- total_eval_batch_size: 16
 - optimizer: Adam with betas=(0.9,0.999) and epsilon=1e-08
 - lr_scheduler_type: cosine
 - lr_scheduler_warmup_ratio: 0.1
-- num_epochs: 15
+- num_epochs: 10
 
 ### Training results
 
-| Training Loss | Epoch | Step | Validation Loss |
-|:-------------:|:-----:|:----:|:---------------:|
-| 0.9333        | 1.0   | 406  | 2.4449          |
-| 0.8399        | 2.0   | 812  | 2.3513          |
-| 0.7898        | 3.0   | 1218 | 2.4311          |
-| 0.7295        | 4.0   | 1624 | 2.5969          |
-| 0.6377        | 5.0   | 2030 | 2.8601          |
-| 0.5891        | 6.0   | 2436 | 3.1033          |
-| 0.5095        | 7.0   | 2842 | 3.5735          |
-| 0.4514        | 8.0   | 3248 | 3.9319          |
-| 0.3872        | 9.0   | 3654 | 4.4386          |
-| 0.3371        | 10.0  | 4060 | 4.8561          |
-| 0.2993        | 11.0  | 4466 | 5.2020          |
-| 0.29          | 12.0  | 4872 | 5.4070          |
-| 0.2802        | 13.0  | 5278 | 5.5084          |
-| 0.2693        | 14.0  | 5684 | 5.5384          |
-| 0.2715        | 15.0  | 6090 | 5.5381          |
+| Training Loss | Epoch  | Step | Validation Loss |
+|:-------------:|:------:|:----:|:---------------:|
+| 1.3554        | 0.9954 | 109  | 2.6451          |
+| 1.0898        | 2.0    | 219  | 2.5083          |
+| 1.0434        | 2.9954 | 328  | 2.4801          |
+| 0.9864        | 4.0    | 438  | 2.4743          |
+| 0.9371        | 4.9954 | 547  | 2.4854          |
+| 0.9157        | 6.0    | 657  | 2.4642          |
+| 0.8657        | 6.9954 | 766  | 2.5076          |
+| 0.8393        | 8.0    | 876  | 2.5159          |
+| 0.8462        | 8.9954 | 985  | 2.5185          |
+| 0.8359        | 9.9543 | 1090 | 2.5157          |
 
 
 ### Framework versions
 
 - PEFT 0.10.0
 - Transformers 4.40.0
-- Pytorch 2.2.1+cu121
+- Pytorch 2.1.2+cu121
 - Datasets 2.18.0
 - Tokenizers 0.19.1

adapter_config.json

@@ -10,23 +10,18 @@
   "layers_pattern": null,
   "layers_to_transform": null,
   "loftq_config": {},
-  "lora_alpha": 16,
+  "lora_alpha": 64,
   "lora_dropout": 0.05,
   "megatron_config": null,
   "megatron_core": "megatron.core",
   "modules_to_save": null,
   "peft_type": "LORA",
-  "r": 8,
+  "r": 32,
   "rank_pattern": {},
   "revision": null,
   "target_modules": [
-    "q_proj",
     "v_proj",
-    "o_proj",
-    "gate_proj",
-    "k_proj",
-    "up_proj",
-    "down_proj"
+    "q_proj"
   ],
   "task_type": "CAUSAL_LM",
   "use_dora": false,

adapter_model.safetensors

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:530dc4e1db6c3022c545725157c6e5d2e7ecf9b878bff1f8f1217168bfc5439e
-size 50056096
+oid sha256:96a298c3833fe7338cd4c0311b9ecf38c8370339602da5926c9ab18098266f0a
+size 25705248

all_results.json

@@ -1,14 +1,14 @@
 {
-    "epoch": 15.0,
-    "eval_loss": 5.538125038146973,
-    "eval_runtime": 2.0562,
+    "epoch": 9.954337899543379,
+    "eval_loss": 2.515676259994507,
+    "eval_runtime": 0.2332,
     "eval_samples": 25,
-    "eval_samples_per_second": 4.863,
-    "eval_steps_per_second": 0.973,
-    "total_flos": 4.655418812370256e+18,
-    "train_loss": 0.9284773617542436,
-    "train_runtime": 39545.4398,
-    "train_samples": 59995,
-    "train_samples_per_second": 2.463,
-    "train_steps_per_second": 0.154
+    "eval_samples_per_second": 42.882,
+    "eval_steps_per_second": 4.288,
+    "total_flos": 3.327732991202951e+18,
+    "train_loss": 2.136770288659892,
+    "train_runtime": 2636.8816,
+    "train_samples": 64610,
+    "train_samples_per_second": 26.554,
+    "train_steps_per_second": 0.413
 }

config.json

@@ -23,9 +23,9 @@
     "_load_in_4bit": true,
     "_load_in_8bit": false,
     "bnb_4bit_compute_dtype": "bfloat16",
-    "bnb_4bit_quant_storage": "bfloat16",
+    "bnb_4bit_quant_storage": "uint8",
     "bnb_4bit_quant_type": "nf4",
-    "bnb_4bit_use_double_quant": true,
+    "bnb_4bit_use_double_quant": false,
     "llm_int8_enable_fp32_cpu_offload": false,
     "llm_int8_has_fp16_weight": false,
     "llm_int8_skip_modules": null,

eval_results.json

@@ -1,8 +1,8 @@
 {
-    "epoch": 15.0,
-    "eval_loss": 5.538125038146973,
-    "eval_runtime": 2.0562,
+    "epoch": 9.954337899543379,
+    "eval_loss": 2.515676259994507,
+    "eval_runtime": 0.2332,
     "eval_samples": 25,
-    "eval_samples_per_second": 4.863,
-    "eval_steps_per_second": 0.973
+    "eval_samples_per_second": 42.882,
+    "eval_steps_per_second": 4.288
 }

runs/Jun13_06-35-32_gpu1-2/events.out.tfevents.1718231867.gpu1-2.1149169.0

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:d4a1cbd13dabfedd2947086fc7c972e092e4e141b4de2b42a321c0f7eb62c0d1
+size 54615

runs/Jun13_06-35-32_gpu1-2/events.out.tfevents.1718234504.gpu1-2.1149169.1

@@ -0,0 +1,3 @@
+version https://git-lfs.github.com/spec/v1
+oid sha256:cc27e99d1c02ca03be4cb57d3d5c9799a043e9730d9316d1740ac833876cbcc9
+size 359

train_results.json

@@ -1,9 +1,9 @@
 {
-    "epoch": 15.0,
-    "total_flos": 4.655418812370256e+18,
-    "train_loss": 0.9284773617542436,
-    "train_runtime": 39545.4398,
-    "train_samples": 59995,
-    "train_samples_per_second": 2.463,
-    "train_steps_per_second": 0.154
+    "epoch": 9.954337899543379,
+    "total_flos": 3.327732991202951e+18,
+    "train_loss": 2.136770288659892,
+    "train_runtime": 2636.8816,
+    "train_samples": 64610,
+    "train_samples_per_second": 26.554,
+    "train_steps_per_second": 0.413
 }

training_args.bin

@@ -1,3 +1,3 @@
 version https://git-lfs.github.com/spec/v1
-oid sha256:086f802354141d938ececcb24962509b618fe3c83ba49547cc844ad2a9563e39
+oid sha256:342607486cb1c55e0124fb386be3828031f9153fde8abaf253c30aa5f8095a8b
 size 5176